How Did Income Affect Education in Tennessee in the 2014 - 2015 Academic Year?

In 2016 ACT reported that students, from families with an annual income of $80,000 or more tend to score about four (4) points higher than students in families that earn less than $80,000 a year.

ACT by Family Income

ACT by Family Income

It’s kindof a big deal

Objective:

Determine how the Achievement Gap can be measured in TN with the publicly-available data from the:

Databases used:

  • Tn. Dept. of Ed. Highschool Achievement Profile Data from the 2014 - 2015 academic year.

  • IRS income-tax-return data from 2013
    • Because 2013 taxes are (expected to be) collected in April 2014, and classes begin in August 2014.

“Wealth Bubbles”

Wealth Distribution Across TN

Using the 2013 IRS Tax Return Data we can plot a bubble for each (inhabited) zipcode in TN. And we can:

  • Make the bubbles larger according to the IRS’s population estimates for 2013

  • Make the bubbles darker according to each zipcode’s Adjuste Gross Income per capita (County AGI divided by County Population).

The results are a bubbles that represent the “wealth concentration” across TN.

  • I.E.
    • A large light bubble represents a large population with a small AGI per capita
    • A small dark bubble represents a small pupulation with a high AGI per capita

Using this data we can represent each county’s Adjusted Gross Income (AGI) as shades of blue.

We achieve this by averaging the AGI of of each zip code within a County.

Here are The average ACT-composite scores for each county.

We avhieve this in a similar fashion by averaqging each districts’s ACT-composite score within each County.

We can see that the “richest” counties do not necessarily have the highest ACT scores.

But, this could be due to multiple factors.

  • Maybe some rich counties have more students
    • So, maybe each student in a rich county actually gets less dollars for their education
  • Maybe some poor counties have fewer students
    • So, maybe each student actually gets more dollars for their education

Drilling down to the granularity of each county’s academic-subject perfomance shows too much variation to be obviously helpful.

School Subjects of the Top 5 Vs. Bottom 5 AGI Per Return

  • To visualize the academic performances of students in wealthier counties, we compared the grade averages of students in the five counties with the highest AGI per tax return against those of the five counties with the lowest AGI per return.

  • Williamson, Knox, and Wilson counties tend to have higher grade averages than the bottom five counties.
    But, Fayette and Davidson counties have high AGI per returns but tend to perform similarly to the bottom five counties.

Let’s Get Statistical

Here is a trés fancy “correlation matrix”

It shows how much of a statistical correlation there is between each of these Key Metrics. Some of them come from IRS tax data, and some of them come from TN’s Dept. of Ed data.

  • agi -> adjusted_gross_income
  • sals -> salaries_and_wages_in_agi_household
  • unemp -> unemployment_compensation_household
  • mort_int -> mortgage_interest_paid_household
  • pct_blk -> pct_black
  • pct_hisp -> pct_hispanic
  • pct_nata -> pct_native_american
  • pct_el -> pct_EL
  • pct_swd -> pct_SWD
  • pct_ed -> pct_ED
  • ppe -> per_pupil_expenditures
  • pct_bhn -> pct_BHN
  • ACT -> ACT_composite
  • pct_abs -> pct_chronically_absent
  • pct_sus -> pct_suspended
  • pct_exp -> pct_expelled
  • grad -> graduation
  • drop -> dropout
  • g_enr -> grade_enrollment
  • g_exp -> grade_expenditure
  • agi_q -> adjusted_gross_income_quantile

In the middle of the diagonal you can find “ACT” and in the bottom left you can find “agi” the correaltion matrix shows that there is actually not a very strong positive correlation between Adjusted Gross Icome (AGI) and ACT scores, but there is a strong negative correlation between ACT scores and the percentage of economically disadvantaged students in a county.

It may be suprising that economically disadvantaged counties have worse ACT scores, but that counties with high AGI do not necessarily have better ACT scores. But that’s because it really depends how you measure the money.

  • We could calculate what percentage of a County’s Income that each student gets (but that’s going to be a very small percent).

  • We could calculate what percentage of a County’s Education-Budget/Expenditure each student gets (not as small, but still pretty small).

  • We could calculate the brute* amount of dollars each student gets.*

  • And Compare them to each county’s ACT score

### Perhaps surprisingly, the brute amount of dollars each student “receives” in a county (a county’s per-pupil expenditure) is not the best predictor of that county’s average ACT score.

County Expenditure

Money distribution

We wanted to determine how a county’s education expenditure compared to its adjusted gross income as well as how that expenditure related to the overall Tennessee education budget. The following graph indicates that counties with a high expenditure-to-AGI ratio typically do not have budgets that make up a significant portion of TN’s overall budget. On the other hand, the counties that tend to have a lower expenditure-to-AGI ratio require more of TN’s total education budget.

Household Wealth And The ACT

County AGI Per Return Impact on ACT Performance

Having heard the rumors of wealthier students performing better on standardized tests, we wanted to put this to the test. In taking the average AGI per return, we had a reasonable idea of the wealth of each household in each county. The plotting of this metric against the average ACT score indicates that a moderately strong correlation exists between the two. In other words, the data suggests that the students in counties with wealthier households tend to score higher on the ACT.

  • enr -> county_enrollment
  • exp -> county_expenditure
  • act -> county_ACT_composite
  • ppe -> county_per_pupil_expenditure
  • agi -> county_adjusted_gross_income
  • num_returns -> county_number_of_returns
  • ppe_agi -> ppe_as_percentage_of_county_agi
  • ppe_exp -> ppe_as_percentage_of_county_expenditure